WatchTower

Overview

WatchTower is a monitoring and deployment platform I built end to end: install an agent on a server, and the dashboard fills with live host, container and security telemetry — then deploy to that server straight from GitHub, or just ask the built-in AI assistant to do it. It is one product covering two jobs that usually need two tools: observability and release management.

Key Features

• Custom Rust monitoring agent compiled to a ~2MB dependency-free static musl binary, cross-compiled through Docker and installed over SSH, streaming host, container and security telemetry from every monitored server on a 10-second interval.
• Real-time observability dashboard rendering 27 live charts across 6 metric groups, with a global alerting layer that aggregates CPU, memory, disk and load-average threshold breaches across all servers into one unified notification panel.
• GitHub-to-cloud deployment pipeline with branch-tracked staging and production releases, one-click deploys and promotions, safe git revert-based rollbacks, AES-encrypted SSH credential storage and automated Slack release notifications.
• AI operations assistant — the full deploy workflow exposed as agent tools, letting users list repositories, launch sandboxes, promote releases and post Slack updates in natural language, with each sandbox auto-onboarded into the monitoring dashboard.
• Full-stack foundation: a React 19 and Tailwind dashboard, a Node.js/Express API with session-based auth, PostgreSQL for persistence, and a Python LLM microservice powering the agent.

Impact

WatchTower collapses the gap between noticing a problem and shipping the fix. The same interface that surfaces a memory alert can roll the offending release back — and because every deploy action is an agent tool, the whole workflow is reachable in plain English.